Stop relying on 'known_events' when adding metrics to Service Ping

As discussed here, we are currently loading Redis events from known_events yaml files and automatically adding their metrics to Service Ping by adding a _weekly and _monthly suffix to their name.

This is happening in the unique_events_data method of HLLRedisCounter.

The same metrics also need to be defined using metrics definitions, creating duplication and room for error [like the one from linked discussion]. To solve this, we should stop using known_events when creating the Service Ping load and just add the metrics based on their instrumentation classes.